Deep Learning: Practical Neural Networks with Java

Summary

In this chapter, we've an application of customer profiling using the Kohonen neural network. Unlike the classification task, the clustering task does not consider the previous knowledge on the desired output; instead it is desirable for the clusters to be found by the neural network. However, we've seen that validation techniques may include external validation, which is a comparison with what could be understood as target output. Customer profiling is important because it gives a business owner more accurate and clean information about their customers, without the human interference in pointing which customers are in some groups or in others, as occurs in supervised learning. That's the advantage of unsupervised learning, enabling the ...